Btrfs: don't make a file partly checksummed through file clone
PATCH 1/2 Btrfs: don't make a file partly checksummed through file clone
To reproduce the bug:
# mount /dev/sda7 /mnt # dd if=/dev/zero of=/mnt/src bs=4K count=1 # umount /mnt
# mount -o nodatasum /dev/sda7 /mnt # dd if=/dev/zero of=/mnt/dst bs=4K count=1 # clone_range -s 4K -l 4K /mnt/src /mnt/dst
# echo 3 > /proc/sys/vm/drop_caches # cat /mnt/dst # dmesg ... btrfs no csum found for inode 258 start 0 btrfs csum failed ino 258 off 0 csum 2566472073 private 0
It's because part of the file is checksummed and the other part is not, and then btrfs will complain checksum is not found when we read the file.
Disallow file clone if src and dst file have different checksum flag, so we ensure a file is completely checksummed or unchecksummed.
--- fs/btrfs/ioctl.c | 5 +++++ 1 files changed, 5 insertions(+), 0 deletions(-)
fs/btrfs/ioctl.c +5 -0
@@ -2177,6 +2177,11 @@ static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,if (!(src_file->f_mode &FMODE_READ )) goto out_fput; +/* don't make the dst file partly checksummed */ +if ((BTRFS_I (src)->flags &BTRFS_INODE_NODATASUM ) != + (BTRFS_I (inode)->flags &BTRFS_INODE_NODATASUM )) + goto out_fput; + ret = -EISDIR ;if (S_ISDIR (src->i_mode) ||S_ISDIR (inode->i_mode)) goto out_fput;
-- 1.7.3.1
PATCH 2/2 Btrfs: don't change inode flag of the dest clone file
The dst file will have the same inode flags with dst file after file clone, and I think it's unexpected.
For example, the dst file will suddenly become immutable after getting some share of data with src file, if the src is immutable.
--- fs/btrfs/ioctl.c | 1 - 1 files changed, 0 insertions(+), 1 deletions(-)
fs/btrfs/ioctl.c +0 -1
@@ -2434,7 +2434,6 @@ static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,if (endoff > inode->i_size)btrfs_i_size_write (inode, endoff); -BTRFS_I (inode)->flags =BTRFS_I (src)->flags; ret =btrfs_update_inode (trans, root, inode);BUG_ON (ret);btrfs_end_transaction (trans, root);
-- 1.7.3.1
Re: [PATCH 2/2] Btrfs: don't change inode flag of the dest clone file
Quoted 5 lines · Li Zefan
The dst file will have the same inode flags with dst file after file clone, and I think it's unexpected.
For example, the dst file will suddenly become immutable after getting some share of data with src file, if the src is immutable.
Good catch! (I did not find any further direct assignment of two inode flags.)
Quoted 2 lines
Signed-off-by Li Zefan <[email protected]>
IMNSHO should go to stable.
david
Quoted 21 lines
--- fs/btrfs/ioctl.c | 1 - 1 files changed, 0 insertions(+), 1 deletions(-)fs/btrfs/ioctl.c +0
-3@@ -2434,7 +2434,6 @@ static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,if (endoff > inode->i_size)btrfs_i_size_write (inode, endoff); -BTRFS_I (inode)->flags =BTRFS_I (src)->flags; ret =btrfs_update_inode (trans, root, inode);BUG_ON (ret);btrfs_end_transaction (trans, root); -- 1.7.3.1 --To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Quoted 17 lines · Li Zefan
To reproduce the bug:
# mount /dev/sda7 /mnt # dd if=/dev/zero of=/mnt/src bs=4K count=1 # umount /mnt# mount -o nodatasum /dev/sda7 /mnt # dd if=/dev/zero of=/mnt/dst bs=4K count=1 # clone_range -s 4K -l 4K /mnt/src /mnt/dst# echo 3 > /proc/sys/vm/drop_caches # cat /mnt/dst # dmesg ... btrfs no csum found for inode 258 start 0 btrfs csum failed ino 258 off 0 csum 2566472073 private 0
Quoted 7 lines
It's because part of the file is checksummed and the other part is not, and then btrfs will complain checksum is not found when we read the file.
Disallow file clone if src and dst file have different checksum flag, so we ensure a file is completely checksummed or unchecksummed.
Signed-off-by Li Zefan <[email protected]>
Looks good to me.
Quoted 23 lines
--- fs/btrfs/ioctl.c | 5 +++++ 1 files changed, 5 insertions(+), 0 deletions(-)fs/btrfs/ioctl.c +5
-1@@ -2177,6 +2177,11 @@ static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,if (!(src_file->f_mode &FMODE_READ )) goto out_fput; +/* don't make the dst file partly checksummed */ +if ((BTRFS_I (src)->flags &BTRFS_INODE_NODATASUM ) != + (BTRFS_I (inode)->flags &BTRFS_INODE_NODATASUM )) + goto out_fput; + ret = -EISDIR ;if (S_ISDIR (src->i_mode) ||S_ISDIR (inode->i_mode)) goto out_fput; -- 1.7.3.1
Quoted 5 lines · Li Zefan
It's because part of the file is checksummed and the other part is not, and then btrfs will complain checksum is not found when we read the file.
Disallow file clone if src and dst file have different checksum flag, so we ensure a file is completely checksummed or unchecksummed.
Your fix prevents the bug, but I don't think it's good to let file clone fail without any other message. ret is set to -EINVAL at the time of 'goto out_fput', which is fine, but the user has no clue what happened or how to fix it.
The nodatasum status is recorded in inode flags and remains like that regardless of the 'mount -o nodatasum', persistent and de facto unchangable (unless the file is created again with the opposite nodatasum mount). Even more, the user has no way to find out nodatasum flag of any inode/file (the corresponding FS_NODATASUM_FL is not there).
My suggestion how to fix this:
1. add FS_NODATASUM_FL file flag and code to set/get via setflags ioctl
2. [this patch to skip cloning in case of nodatasum flag mismatch]
3. ... add a printk why it failed
The user then has at least option to drop/add the nodatasum flag for one of the. Unfortunatelly this makes file cloning less straightforward.
david
Quoted 27 lines
Signed-off-by Li Zefan <[email protected]>--- fs/btrfs/ioctl.c | 5 +++++ 1 files changed, 5 insertions(+), 0 deletions(-)fs/btrfs/ioctl.c +5
-2@@ -2177,6 +2177,11 @@ static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,if (!(src_file->f_mode &FMODE_READ )) goto out_fput; +/* don't make the dst file partly checksummed */ +if ((BTRFS_I (src)->flags &BTRFS_INODE_NODATASUM ) != + (BTRFS_I (inode)->flags &BTRFS_INODE_NODATASUM )) + goto out_fput; + ret = -EISDIR ;if (S_ISDIR (src->i_mode) ||S_ISDIR (inode->i_mode)) goto out_fput; -- 1.7.3.1 --To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Quoted 11 lines · David Sterba
Quoted 5 lines · Li Zefan
It's because part of the file is checksummed and the other part is not, and then btrfs will complain checksum is not found when we read the file.
Disallow file clone if src and dst file have different checksum flag, so we ensure a file is completely checksummed or unchecksummed.
Your fix prevents the bug, but I don't think it's good to let file clone fail without any other message. ret is set to
-EINVALat the time of 'gotoout_fput', which is fine, but the user has no clue what happened or how to fix it.
While I agree with you on this comment..
Quoted 9 lines
The nodatasum status is recorded in inode flags and remains like that regardless of the 'mount -o nodatasum', persistent and de facto unchangable (unless the file is created again with the opposite nodatasum mount). Even more, the user has no way to find out nodatasum flag of any inode/file (the corresponding
FS_NODATASUM_FLis not there).My suggestion how to fix this:
1. addFS_NODATASUM_FLfile flag and code to set/get via setflags ioctl
This means we can have a file partly checksummed, which is what we want to avoid in this patch.
Quoted 2 lines
2. [this patch to skip cloning in case of nodatasum flag mismatch]
3. ... add a printk why it failed
I don't think this is a good idea.
Quoted 4 lines
The user then has at least option to drop/add the nodatasum flag for one of the. Unfortunatelly this makes file cloning less straightforward.
I don't know if Chris has plan on finer-grained checksum (not per file but per extent), if yes, we can eliminate this constraint in file cloning in the future.